Accident prevention in agricultural machinery using radiofrequency technology: A prototype application to olive net collectors

Accident prevention in agricultural machinery using radiofrequency technology: A prototype application to olive net collectors Fernando Chac´ on a , Antonio Cubero-Atienza a , Francisco V´ azquez b , Juan Garrido b , Mario L. Ruz c,* a Department of Rural Engineering, Civil Constructions and Project Engineering, Universidad de C´ ordoba, Campus de Rabanales, 14071 C´ ordoba, Spain b Department of Electrical Engineering and Automatic Control, Universidad de C´ ordoba, Campus de Rabanales, 14071 C´ ordoba, Spain c Department of Mechanics, Universidad de C´ ordoba, Campus de Rabanales, 14071 C´ ordoba, Spain ARTICLE INFO Keywords: Olive net collectors IoT Radiofrequency Machine safety Agricultural safety ABSTRACT Machine related injuries and fatalities are a major concern across industries. Furthermore, machine safety in the agriculture sector is critical due to the high-risk nature of agricultural work, which often involves heavy machinery and equipment. This study investigates radiofrequency systems designed to improve worker safety when operating olive net collectors and similar machinery. An RFID prototype safety device with IoT capabilities was developed and tested. The study also provides an overview of inductive coupling antennas, including a design methodology that consider various shapes. MATLAB code is provided as supplementary material to this paper for calculating inductance, a key parameter of inductive coupling antennas. A custom 3D antenna based on a rectangular shape was designed an integrated with an olive net collector. The entire system was evaluated using a robotic approach, and a set of statistics was obtained to assess its performance. The conducted tests demonstrate that the developed prototype effectively sends an emergency stop signal and serves as a safety barrier, highlighting its potential benefits. 1. Introduction Machine related injuries and fatalities are a major concern across industries. Machine safety is a primary interest, and implementing robust safety protocols and innovative devices not only protects employees from potential hazards but also enhances operational efficiency by reducing the likelihood of accidents that can cause significant interruptions. In addition, innovative machines are constantly developed, and safety aspects must be considered. The agriculture sector is experiencing significant advancements in machinery automation, and the next generation of agroindustry systems will exploit the digitalization technologies, such as the Internet of Things, robotics, Artificial Intelligence and Cloud Computing (Aiello et al., 2022). There is a trend towards the development of intelligent and efficient machines that can assist humans in labor-intensive and repetitive tasks, being one of such advancements the task of harvesting (Kaur et al., 2023). These developments should include improved safety devices to increase operator safety, including adequate consideration of human factors and human machine interaction (Vigoroso et al., 2025). In the European Union, a total of 3347 work-related fatalities were recorded in 2021, and nearly two-thirds of these cases occurred in industries where humans interact with off-road machinery (Aiello et al., 2022). Following the last statistical data of EU, Spain is above the media in mortal accidents at work (Eurostat, 2023). Furthermore, machine safety in the agriculture sector is critical due to the high-risk nature of agricultural work, which often involves heavy machinery and equipment (Mucci et al., 2020). Accidents in the agriculture sector are worth studying, (Ouattara et al., 2023). In fact, it is one of the economic sectors with the highest incidence rate (Fargnoli et al., 2018). Baraza and Cuguer´ o (2021) demonstrated that the fatal accident rate in the agricultural sector between 2016 and 2018 was 60 %–70 % higher than the average across other sectors, while the total accident rate remained similar. An exhaustive statistical analysis of the agricultural sector is also provided by Baraza, 2021. According to data provided by Eurostat 2022 (extracted in october 2024), the incidence rate of non-fatal accidents in Spain was the third highest in the European Union, following Portugal and France (Fig. 1), (Eurostat, 2023). The same Eurostat dataset, categorized according to * Corresponding author. E-mail address: [email protected] (M.L. Ruz). Contents lists available at ScienceDirect Safety Science journal homepage: www.elsevier.com/locate/safety https://doi.org/10.1016/j.ssci.2025.106875 Received 17 October 2024; Received in revised form 9 April 2025; Accepted 10 April 2025 Safety Science 188 (2025) 106875 0925-7535/© 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/bync-nd/4.0/ ). the ‘Nomenclature statistique des Activit´ es ´ economiques dans la Communaut´ e Europ´ eenne’ (NACE), indicates that the Agriculture, Fishing and Forestry sector presents a significantly higher fatal accident rate compared to non-fatal accidents (Fig. 2). Considering that Andalusia is the European region with the largest olive tree surface area in the world, with 1.638.320 ha in 2021 (Anuario de Estadísticas Agrarias y Pesqueras de Andalucía 2021 (CAPADR)), and that olive groves represent 49.12 % of the farmland in Andalusia, according to data from the Spanish Ministry of Agricultural, Phishing and Feeding “Encuesta sobre Superficies y Rendimientos de Cultivos (ESYRCE 2024)” (ESYRCE, 2024), it is evident that preventing accidents involving net collectors, one of the key pieces of equipment used during fruit harvesting, is of critical importance in this sector. Aware of this reality, the authorities in Andalusia have launched highly specific campaigns. Thus, the Employment, Business, and Self-Employment delegation initiated a program in the 2023–24 campaign aimed at improving information and raising awareness, titled ”uso seguro del recogedor de fardos“ (safe use of the net collector), (Baena Hoy, 2023). Given these concerns and the lack of effective safety mechanisms in net collectors, this study aims to develop a system that improves worker protection. Within the different types of machinery used in the agricultural sector, innovation in machine harvesting offers a promising solution, with the potential to reduce harvestings costs and contribute to overall production efficiency (Kaur et al., 2023). Particularly in Spain, which produces a high proportion of fruit and vegetable crops (Baraza and Cuguer´ o-Escofet, 2021), new harvesting methods are constantly proposed, such as olive harvesters, olive net collectors and trunk shakers (Sola-Guirado et al., 2023), (Homayouni et al., 2023). This type of machinery has become established in most olive-growing farms, facilitating the work of collecting and transporting the fruits, saving time, and reducing load handling. One of the most common accidents occurs when a worker is situated within the machine’s operational radius and the machine operator initiates a movement (Gizlenci and Aybek, 2021). In the event that the operator is unable to visually ascertain the presence of workers within the operational radius, an accident may occur, potentially resulting in severe injury. In addition, the use of this type of machinery implies that situations in which a worker needs to handle parts close to the operation region, and the use of standard guards or devices (e.g. physical barriers) is infeasible. In many cases, the upper limb and hand are often the body parts that sustain most damage as there are mostly involved in driving tool and close to moving parts of the machine (Mucci et al., 2020). For this casuistry, a “no-hands-in-die” policy from the point of operation is applied, maintaining a safe distance from the operation point (ISO 13857: 2019). However, given these constraints, in the case of an accident the worker must participate in the activation of an emergency stop, and the stop is often activated after a serious injury has occurred. Olive net collectors belong to this category of machinery and can be dangerous in a similar way to press brakes and cutting machines (L´ opezGallego et al., 2018). In addition, the minimization of these risks depends to a large extent on the training of both the machine operator and the workers who may be in the vicinity of the machine. Within the context of industrial safety, several studies in recent years Fig. 1. Non-fatal accidents that imply at least 4 full calendar days of albescence from work (Eurostat, 2023). F. Chac´ on et al. Safety Science 188 (2025) 106875 2 have explored the use of RFID systems for safety applications. For example, Motroni et al. (2021) proposed a passive RFID-based ranging system to enhance worker safety in agricultural and forestry environments where remote-controlled farm machinery and human workers operate simultaneously, although this study does not provide statistical accuracy metrics. A subsequent work by the same authors (Bandini et al., 2023) incorporated a 3D cartesian portal which exploited synthetic aperture radar (SAR) based localization, reporting a standard deviation of the localization error of 12.5 cm and 17.7 cm of eight UHF tags located within a volume of 420 cm ×165 cm ×40 cm. In Zhang et al. (2023a), a combination of RFID and IoT systems was employed to detect entry and exit activities when workers access a specific virtual zone. However, no statistical data were provided regarding detection distances. Other works have developed proximity detection systems based on magnetic field generators. For example, Li et al. (2012) proposed a magnetic proximity detection system, specifically developed for mining vehicles and mobile machinery to protect nearby workers. While this work was primarily based on the development of a theoretical model, an experimental campaign was also carried out where two magnetic field generators were positioned 1.0 m. Using the measurements from the two alternately pulsed fields, the positions of 40 points on a triangular shaped path were determined. Considering these 40 Fig. 2. Fatal and non-fatal accidents at work by NACE Section (Eurostat, 2023). F. Chac´ on et al. Safety Science 188 (2025) 106875 3 points, the mean location error was 6.58 mm, the standard deviation of the error 2.27 mm, and the maximum error 11.67 mm compared to the actual location measurements. In another work from the same authors (Li et al., 2019), the distortion in the magnetic field distribution caused by the presence of metallic objects was analyzed. While the aforementioned studies are related to the present research, they focus on different design objectives and protection requirements. To the authors’ knowledge, no automatic stop method has been developed to control the risk of entrapment in olive net collectors. The prototype developed in this research offers a solution that proves effective for this scenario. This research is primarily focused on proposing safety systems capable of mitigating various types of accidents involving agricultural machinery, where workers sustain injuries due to contact with unprotected moving parts. To the best of our knowledge, we have not found works that evaluate protective equipment with the capacity to distinguish whether the object entering a dangerous area is a workpiece or the body part of a worker. There are other technologies, including Ultra Wide Band (UWB), computer vision or inertial measurement units (IMUs) (Zhang et al., 2023b; Ruiz et al., 2024), that could potentially provide benefits in combination with the prototype proposed in this work. In this paper, we investigate the use of radiofrequency systems to improve agricultural safety operating machinery similar to olive net collectors. The remaining part of the work is organized as follows: Section 2 illustrates the overall methodology of the developed safety system, including the use of olive net collectors and the primary risks at the operation point. Furthermore, an overview of radiofrequency systems based on inductive coupling is provided, along with a detailed methodology for developing custom inductive coupling antennas. The section concludes with a description of the experimental benchmark employed, which is based on a robotic approach. Section 3 presents and discusses the experimental tests conducted to evaluate the robustness of the developed prototype in performing an emergency stop. The experimental results are discussed, reporting the prototype’s effectiveness as well as the current limitations. Finally, Section 4 summarizes this research work, discusses its limitations and suggests directions for future research. 2. Background, materials and methods 2.1. Net collectors risk analysis A net collector system comprises the deployment of bales or blankets, specifically positioned for this purpose, to facilitate the descent of the olives. Subsequently, a box equipped with motorized rollers is employed to facilitate the collection and loading of the fruit. In certain instances, the operator may encounter difficulties when attempting to grasp the extremities of the bundle, as illustrated in Fig. 3. This maneuver has been identified as a potential cause of accidents due to the risk of the worker’s arm becoming caught in the dragging mechanisms. In addition, the fact that many manufacturers affix the CE marking may increase the risk by giving a false sense of safety. In the southern Spanish region of Andalusia, at least 10 accidents of this type occur every year (Junta de Andalucía. Consejería de Empleo Empresa y Trabajo, 2023). Despite the implementation of measures to mitigate this risk, such as reducing the speed of the moving parts, or including physical barriers, serious accidents still occur. Net collectors present a high risk of entrapment due to the lack of a guard to prevent access to the risky area. The procedure followed by manufacturers to obtain the CE mark complies with article 12 of Directive 2006/42/CE (EC, 2006), specifically through self-certification via Assessment of conformity with internal checks on the manufacture of machinery (Annex VIII). It is important to note that, after conducting a risk assessment and analyzing the state of the art, no effective protective solution currently exists to control or eliminate the risk of entrapment between the wheels. In such cases, Directive 2006/42/EC allows manufacturers to provide adequate information to users regarding residual risks and the necessary safety measures they should take. It can be reasonably argued that the occurrence of an accident is largely dependent on the actions of the operator. Every farmer with such a machine should be aware of this risky situation and contact the manufacturer to implement the essential safety requirements. A tree of causes of this type of accident is shown in Fig. 4. One of the causes highlighted in the figure is the absence of automated protective devices. Fig. 3. (a) Direction in which the motorized wheel system normally turns when the bundle passes, (b) Area with risk of entrapment, (c) Situation in which the accident might happen. Adapted from (Junta de Andalucía, 2023). F. Chac´ on et al. Safety Science 188 (2025) 106875 4 However, current protective devices consist of physical barriers that can improve worker safety, but as mentioned above, the operator must intervene in the emergency stop. All of this highlights that, at present, there is no system that can be considered safe enough to prevent accidents in all cases. A system is required that combines the operator’s ability to act in the event of a mechanism jam with the cessation of moving elements when any part of the operator’s body (hand, arm, etc.) enters the designated danger zone. The system developed here successfully integrates both requirements. This would ensure compliance with the guidelines set forth in the European Directive 2006/42/CE concerning the intrinsic safety of the machine’s operating mode. 2.2. Overview of inductive coupling systems and RFID safety system prototype In the field of telecommunications, RFID is a broad term encompassing any technology that utilizes radio waves to identify, monitor, and thereby manage various objects. A basic RFID system basically consists mainly of the following three components: RFID tags, a reader, and the reader antenna. RFID tags (also known as transponders) are small devices that use radio waves to transmit data to the reader. They can be classified as passive, active, or semi-passive. Passive tags rely on the electromagnetic field from the reader for power, while active tags have their own battery, and semi-passive tags combine both features. RFID readers, can be fixed or mobile, emit radio waves and receive signals from the tags. Both tags and reader use antennas to enable communication. The reader antenna transmits radio waves, and the tag antenna captures these waves to power the tag and transmit data. Additionally, software for managing and filtering the data collected by the reader is indispensable, as is a power source for the reader and its antenna. This section describes the main components of an RFID system based on inductive coupling. Specifically, we focus on low-frequency systems (125–134.5 kHz) and inductive coupling since this method is less affected by metals on the surroundings (Li et al., 2012). First, the main characteristics of inductively coupled antennas are described. Second, the RFID system prototype is described. The protective equipment described in this work can be categorized as an electrosensitive barrier that differentiates between the net of olives or the hands of a worker, which have a transmitter attached. In the following, we focus on the reader antenna, which is a critical component responsible for covering the dangerous region. We provide a basic guideline for building and tuning custom inductive coupling antennas. 2.2.1. Inductive coupling antennas The RFID reader antenna is responsible for receiving data from the tags and supplying them the energy needed to transmit data. Lowfrequency RFID antennas are composed of a looped wire that creates an oscillating magnetic field when powered by and alternating power source. Passive RFID tags operate using an induced voltage from the antenna coil. By emitting an energizing RFID signal, a reader can interact with a remotely located device that lacks an external power source, such as the tags used in this work. This interaction and communication occur through antenna coils, making it crucial for the device to have a properly designed antenna circuit for effective RFID applications (Microchip Technology, 2003). Low-frequency antennas transfer energy to the transponder via the magnetic field. When current flows through an antenna, magnetic field lines are generated like those depicted in Fig. 5a. When a tag and the reader antenna are in close proximity, the time-varying magnetic field produced by the reader antenna coil induces a voltage in the tag antenna coil. This induced voltage causes a flow current on the coil. The detection distance is influenced by the orientation of the tag relative to the antenna, resulting in the field patterns shown in Fig. 5b. Custom-built antennas are often essential for various reasons (Texas Instruments, 2003). In typical RFID applications, antennas must be integrated into structures or equipment, such as doors, large-scale antennas like road loops, or compact antennas for localized reading. The reader antenna typically forms a series or a parallel resonant circuit and must be tuned to the operating frequency to maximize power efficiency. The following formula determines at what frequency the antenna circuit resonates: fresonance =1 2 π  LC √,(1) where L (H) represents the inductance of the antenna circuit and C (F) the capacitance. The employed RFID module in this work operates at 134.2 kHz and thus the resonant circuit must match fresonance (Hz). This is carried out either by varying the antenna inductance L or adding a specific capacitance C. For general applications, this circuit is tunned so that the maximum current circulates, and thus a magnetic field of greater intensity is generated. If the antenna is not tuned, the reading Fig. 4. Accident cause tree, adapted from (Junta de Andalucía. Consejería de Empleo, Empresa y Trabajo, 2023). F. Chac´ on et al. Safety Science 188 (2025) 106875 5 range will decrease (i.e., the tags need to be closer to the reader antenna to be able to send their information). Thus, estimating the reader antenna inductance is critical for an optimal operation of the system. This parameter mainly depends on physical variables such as antenna height, width and shape geometry and must be within a specific range depending on the reader to allow the process of fine-tuning. For relatively simple planar geometries, analytical expressions for calculating the inductance are available in the literature (Grover, 1946; Welsby, 1960; Microchip Technology, 2003). The inductance also depends on the wire size, the number of turns, and the spacing between them. Fig. 6 illustrates four simple planar shapes that could be useful in dangerous areas of small size, and Table 1 shows the mathematical expressions for their calculation, where N represents the number of turns and the remaining parameters are the antenna dimensions in centimeters. The MATLAB code for the inductance calculation of different planar shapes is provided as supplementary material to this paper and is available in the GitHub and Zenodo repositories. It should be noted that the analytic inductance estimation serves as a starting point and a fineFig. 5. a) Magnetic field lines (left), b) detection patterns depending on tag orientation. Fig. 6. Inductive coupling antennas of planar shapes. F. Chac´ on et al. Safety Science 188 (2025) 106875 6 tuning step is needed. Additionally, the antenna tail must be also considered as it contributes to increase the inductance. In the specific case of the RFID radiofrequency module used in this study, an antenna inductance of 27 μ H is required to reach the resonance point. For example, a rectangular antenna with dimensions of 0.28 m in heigh and 0.34 m width requires 6 turns to achieve an inductance value of the 28.94 μ H, being this one the closest to 27 μ H. The inductance value as a function of the number of turns is shown in Fig. 7. Another important aspect for the antenna inductance value is the presence of metals in the surroundings, as this will change the antenna inductance significantly. This last factor is crucial in the presented casuistry since the antenna is attached to a machine, and thus the presence of different metal materials in the nearby are expected. Detuning may occur when the antenna is mounted close to metal (the inductance decreases) or when a long tail is added (the inductance increases). Metal close to the antenna or a large mass of metal relative to the antenna size can decrease the antenna performance. The tuning procedure implies adding external capacitors in series or parallel if detuning occurs. High voltage polypropylene capacitors are used for this purpose (Texas Instruments, 2003) or even changing the shape of the loop. On the other hand, the behavior of AC current through a wire is different from the flow through a DC circuit. The so-called “skin effect” must be considered, since it increases the wire resistances in AC circuits. Therefore, the type of wire must be considered. As mentioned, the RFID reader utilized in this research operates at 134.2 kHz, requiring that any antenna be accurately calibrated to achieve resonance at that frequency. The flow diagram in Fig. 8 illustrates the basic steps for antenna tuning. The fine-tuning step must be carried out once the antenna has been installed. The total inductance of the antenna can be measured with an inductance meter. In addition to ensuring operation at the resonance point (i.e., matching 27 μ H for our RFID module), the power supply will directly affect to the current flowing through the antenna. This characteristic is utilized in this work to regulate the maximum detection distance, although the orientation of the tag is also important. Finally, another important parameter is the quality factor of the antenna Q, which is a measure of its effectiveness. Typical Q values with the RFID module used are within the range 60–200 and can be controlled selecting the wire type (Texas Instruments, 2003). It must be noted that a higher Q value increases the likelihood of the antenna being de-tuned by proximity to metal. In addition to the reader antenna characteristics (size and shape), the reading distance is influenced by other factors, including the size and shape of the transponder antenna, the level of electrical noise in the environment and the power of the transmitter. 2.3. RFID safety system prototype overview As previously explained, it is essential to break the chain of accident occurrence. Therefore, the system design follows a reverse engineering approach. Specifically, based on the identified causes of accidents (extracted from databases containing records of this type of incidents), a prototype system has been designed to interrupt the sequence of events leading to an accident, thereby preventing its occurrence. This methodology has been successfully employed in prior research, yielding effective results (Gattamelata et al., 2023, 2021). By analyzing the accident mechanisms described in investigations involving this type of machinery, the most effective method for breaking the accident chain has been identified, making this reverse engineering approach both practical and efficient in mitigating potential accidents. To establish the reverse engineering process, the following factors were taken into account: – Conducting an analysis of how operators approach the point of operation, with ergonomic aspects considered in the design of the proposed system. –Incorporating the real-world experience of operators and manufacturers of this type of machinery, including several interviews with ALUA Agricultural Machinery, one of the largest manufacturers of olive net collectors. – Developing an RFID-based tag detection system for identifying proximity to the danger point, utilizing an antenna and adhering to all relevant international standards. – Validating the system at the laboratory level as a preliminary step, using an industrial robot to simulate approach trajectories to the point of operation. Table 1 Analytical expressions for inductance calculation of the antenna shapes shown in Fig. 6. All dimensions must be introduced in cm, L is in μ H. Antenna type Inductance estimation, L ( μ H) Multilayer circular coil L=0.31(aN)2 6a+9h+10b Spiral coil L=0.397(aN)2 8a+11b Multilayer square coil L=0.008a0.397(aN)2 8a+11b Multilayer rectangular coil L=0.0276(CN)2 1.908C+9b+10h Fig. 7. Inductance value of a 0.34 ×0.28 m rectangular antenna as a function of the number of turns. F. Chac´ on et al. Safety Science 188 (2025) 106875 7 The RFID safety system prototype is detailed below. Fig. 9 illustrates a basic scheme of the developed safety device. Components include the RFID reader itself as well as a printed circuit board with a microcontroller that sets up the reader via a RS232 interface, activates a digital output that carries out the emergency stop with the possibility of sending a LoRa-based signal transmission. The procedure is as follows: once the tag (i.e., an RFID transmitter that would be embedded in the wristband of the operator) is detected by the antenna, the controller activates one of its digital outputs stopping the machine and sending a message via LoRa wireless protocol. In addition, the RFID module has an input for the antenna connection. In this case, the designed antenna has a 3D shape specifically developed to be assembled in the dangerous zone of an olive collector prototype. A set of condensers and a regulable inductance (not shown in Fig. 9) are used to set the resonance point once the antenna is installed. As mentioned in the previous section, the correct design reader antenna is critical for the proposed application and ensuring a correct tuning is critical. The power supply submodule is responsible for supply power to both the microcontroller and the RFID module. As will be demonstrated in Section 3, the applied voltage has an impact on the distance range. 2.3.1. Olive net collector and antenna coupling A prototype olive collector and a 3D inductive coupling antenna were designed specifically for experimental testing. The RFID antenna developed covers a front and rear area of approximately 0.44 m wide and 0.25 m high. The lateral areas are 0.45 m wide with an average height of 0.22 m high (the frontal side has a height of 0.45 m and the rear side 0.2 m). The risky area of the olive net collector is the place where the operator has to introduce the net. In this case, the volume covered by the antenna is approximately 0.0445 m 3 , covering the operation point. To estimate the inductance of the antenna, the procedure explained in Section 2.2 was used. The inductance estimation was obtained first assuming that the 3D shape shown in Fig. 10a is a flat multilayer rectangular coil (4th equation in Table 1) of 0.4 m x 0.93 m. This gave a theoretical inductance of 35 μ H for 5 turns. This value will decrease since the antenna is installed close to metal and front and rear parts are bent. A polypropylene capacitor of 22 μ F was added in series and the fine-tuning procedure, based on a trial-and-error approach, was carried out to reach the resonance point. The type of wire used was ‘oxygen-free’ of a section of 1.5 mm 2 . The frame was printed with polylactic acid (PLA). Fig. 10b shows the prototype frame and the antenna installed in the olive net collector. Fig. 8. Phases for the design and installation of inductive coupling antennas. Fig. 9. Schematic representation of the developed safety device. F. Chac´ on et al. Safety Science 188 (2025) 106875 8 2.4. Design of experimental testing The designed antenna was tested using the robotic approach described in (Ruz et al., 2012). This approach was employed to calculate independent statistics for each entry point, a crucial requirement due to the fact that metal parts and varying working environments can alter the shape of the detection zones. The ABB IRB 2400L controller allows the robot’s end effector to be positioned with an accuracy of 1 mm (ABB Robotics AB, a). A set of points are defined through which the end effector moves, creating a set of linear paths as illustrated in Fig. 11a. The stored detection distances per entry point are reorganized in matrices. Each element of the matrix indicated a unique location for the occurrence of each measurement/value. As previously stated, in the case of the olive net collector, the dangerous region is the one surrounding the wheels, specifically the part where the net enters. RobotStudio (ABB, 2024) was used for the definition of the dangerous area. For the tests (Fig. 11b), the RFID microcontroller was programmed so that it sends a digital signal to the robot controller and the position of tag is stored. The microcontroller was programmed to hold the output a number of seconds so that the robot gets out of the detection zone. This procedure is repeated through all the entry points defined in the robot program. The robot pseudocode is shown in Table 2. The generated text file is postprocessed and statistical data are generated. As shown in Fig. 12, each entry point is localized with respect to the antenna. 3. Testing and evaluation A set of twelve experiments were carried out varying the voltage input power applied to the RFID module and the linear velocity of the extreme part of the robot. The modification of the input voltage allowed us to get a measurement of the sensitivity for the distance detection. In all the experiments the antenna worked at its resonance point. The setup simulates the approach of the worker to the dangerous area. When the tag is read by the antenna, the RFID prototype sends a signal to the robot controller and the tag position is stored. This experiment is repeated automatically, and the robot initiates another movement, entering the dangerous area by a different route, as shown in Fig. 11a. The analysis focuses on the plane of operation where the accident can occur, specifically the frontal plane where the net is collected (Fig. 10b and Fig. 12). 3.1. Experimental tests The detection distances were measured from a coordinate system whose origin was located at the point of operation, as illustrated in Fig. 10b. A total of 36 points organized in 4 files of 9 points were preprogrammed covering the frontal area. In each experiment, the extreme part of the robot entered each of the pre-programmed locations at least 10 times, gathering sufficient statistics per point, which resulted in a total of at least 360 distance detections recorded per experiment. Table 3 summarizes the specific details of each experiment. The first column lists the experiment name, the second column indicates the linear velocity of the extreme part of the robot upon entering the hazardous zone, and the third column indicates the applied input voltage on the RFID prototype. Two speeds and five different voltages were tested, where the robot entered through the frontal region, which is the critical region due to the movement of the wheels as shown in Fig. 3c. The entry velocities were 250 mm/s and 500 mm/s, with input voltages ranging from 7 to 18 V, which are the maximum and minimum values that significantly affect the detection distance. Passive RFID tags were mounted in the extreme part of the robot. 3.2. Experimental results Statistical values were obtained for each of the experiments. Fig. 13 illustrates the mean values as a function of the entry points for experiments 1, 2, 3 and 4, in which the linear velocity when entering the risky area was 250 mm/s and the supply voltage increased gradually from 7 to 15 V. The central values, 5th column of the matrices, indicate the point Fig. 10. a) Antenna shape, b) antenna prototype installed at the point of operation. Fig. 11. a) Simulation of trajectories, b) experimental test with antenna prototype and olive net collector. F. Chac´ on et al. Safety Science 188 (2025) 106875 9